Okay, well, I've been able to bring on the fault condition by simply rotating the datum dial whilst in the 'stop frequency' mode, so dirty keys are now looking increasingly unlikely (though they will require attention).
Since the consensus (such as it is!) is to investigate the PSU, I'll do that next. I'll scope the power rails of the individual low voltage supplies out of the main PSU and check their voltage levels are within spec. That should trap 99% of PSU problems. I'll also check the ESRs of the larger electrolytics and any suspicious bulges whilst I'm at it. Anything else I should look at?
Jinxie